Black and Latinx Primary Caregiver Considerations for Developing and Implementing a Machine Learning–Based Model for Detecting Child Abuse and Neglect With Implications for Racial Bias Reduction: Qualitative Interview Study With Primary Caregivers

Editorial reading guide · Respect

Whose understanding of our family enters an AI risk label?

Mothers described language barriers, ambiguous records, and fear of child protective services that could shape an abuse-detection model’s inputs and harms. Respect here begins with listening to communities before building a predictor.

Study context

Twenty Black and Latinx mothers whose children attended a northeastern US pediatric emergency department discussed abuse, neglect, documentation, and child-protection experiences.

Age / stage: Children attending a pediatric emergency department; exact child ages not specified

Read with three questions

1. How might communication problems become misleading training data?
2. What do caregivers’ experiences add beyond a statistical fairness metric?
3. How could model development include families without increasing surveillance harms?

Keep the limits in view

The study used one urban emergency department, English-proficient participants, and mothers only. It informed future model design; it did not build a detector or demonstrate reduced racial bias or safer reporting.